The Iterative Refinement Loop

You've sent your research brief, received the first draft. Now what? Most people either accept it or start over. The better approach: targeted refinement.

The refinement loop has three moves:

1. Zoom In — "Expand on finding #3 with more detail and evidence." 2. Challenge — "What would a skeptic say about this conclusion? What evidence contradicts it?" 3. Restructure — "Reorganize this as a decision brief instead of an executive summary."

## Zoom In
"Finding #3 is the most important. Give me 3x more detail
on that point specifically, including supporting evidence
and counterarguments."

## Challenge
"Play devil's advocate on your top recommendation. What
are the strongest arguments against it? What am I missing?"

## Restructure
"Good analysis, wrong format. Rewrite as a comparison
matrix with these 5 dimensions as columns: [list them]."

The 2-pass refinement pattern works well for high-stakes research: Pass 1 generates the initial analysis. Pass 2 asks Claude to critique its own output, identify weak spots, and strengthen them. The second pass catches hedging language, unsupported claims, and gaps in logic.

Typically, 2-3 refinement rounds are enough. After that, you hit diminishing returns — the output stabilizes and further prompting just rearranges existing content.